Quaternion Convolutional Neural Networks

TL;DR

This paper introduces Quaternion CNNs that leverage quaternion algebra to better capture color image features, outperforming traditional CNNs in classification and denoising tasks.

Contribution

The paper develops a novel quaternion CNN framework with redesigned modules compatible with existing deep learning techniques, enhancing feature representation for color images.

Findings

Outperforms real-valued CNNs in classification accuracy

Achieves better denoising results on color images

Modules are compatible with standard deep learning architectures

Abstract

Neural networks in the real domain have been studied for a long time and achieved promising results in many vision tasks for recent years. However, the extensions of the neural network models in other number fields and their potential applications are not fully-investigated yet. Focusing on color images, which can be naturally represented as quaternion matrices, we propose a quaternion convolutional neural network (QCNN) model to obtain more representative features. In particular, we redesign the basic modules like convolution layer and fully-connected layer in the quaternion domain, which can be used to establish fully-quaternion convolutional neural networks. Moreover, these modules are compatible with almost all deep learning techniques and can be plugged into traditional CNNs easily. We test our QCNN models in both color image classification and denoising tasks. Experimental results show that they outperform the real-valued CNNs with same structures.